Trtllm fp4 block scale MoE topk1 e16 h5120 i8192
0 eligible runs
moe
FP4 block scale MoE (MxFP4+BF16, TRT-LLM style). Llama 4 Scout 17B-16E. Llama4 routing (sigmoid+top-1). Routing: Llama4 (sigmoid+top-k, routing_method_type=3).
Current records
No published measurement for the selected workload.
Implementations
Implementation
Runtime
Best latency
Evidence
Availability
Semantics
Inputs and outputs
routing_logitsbf16 [seq_len, num_experts]
hidden_statesbf16 [seq_len, hidden_size]
gemm1_weightsfp32 [num_experts, gemm1_out_size, hidden_size]
gemm2_weightsfp32 [num_experts, hidden_size, intermediate_size]
routing_biasbf16 [num_experts]
outputbf16 [seq_len, hidden_size]
Axes and behavior
top_kconstant = 1
seq_lenvariable
hidden_sizeconstant = 5120
num_expertsconstant = 16
gemm1_out_sizeconstant = 16384
intermediate_sizeconstant = 8192
determinismunspecified
constraintsNo mutation or aliasing
Identity
No source imports for this operation yet.How records are decidedJSON